Osaka Urban Area Last-Mile Delivery Optimization Challenge 2026
openRules
The submission must be a ZIP file of 25MB or less and include manifest.json and the declared agent. The agent will be executed in an isolated environment with no network connection, a maximum execution time of 300 seconds, 2048MB of memory, and a CPU limit of 2 cores. No additional packages will be installed at runtime. Please use Python, NumPy, SciPy, NetworkX, OR-Tools, Qiskit, and Qiskit Aer, which are included in the fixed sandbox. Output result.json, qubo.json, and quantum.json as valid JSON objects of 5MB or less each. Every delivery destination must be visited exactly once, and only registered vehicles may be used. Vehicle capacities and time windows for each destination must be respected. QUBO variables, coefficients, and penalties must be finite values in a consistent format, and the quantum circuit, number of measurements, measurement results, and selected bitstrings must be submitted in a verifiable format. The random seed for evaluation is fixed at 20260803. Apply QVARENA_SEED to all random number generators used. Evaluation settings and official datasets will be locked after the first submission. In the event of a tie, the ranking will be determined by the number of constraint violations, total distance, execution time, and submission date/time, in that order.
Evaluation
The total score is a weighted sum of the delivery route optimization score (60%) and the quantum score (40%). The delivery score is calculated from the total distance relative to the official baseline_cost, delivery completion rate, and the number of violations regarding capacity, time windows, vehicles, and duplicate deliveries. The quantum score consists of 30 points for QUBO structure, 40 points for verifiable circuits and measurement trails, 20 points for quantum resource efficiency, and 10 points for the correspondence between the selected bitstring and the delivery solution. Circuits will be re-analyzed using Qiskit, and quantum utilization based solely on self-reporting is capped at 40 points. If the delivery does not cover all locations or has non-zero violations, no quantum bonus points will be awarded.
Scoring configuration
Route
60%Route quality relative to the official dataset baseline, adjusted for delivery coverage and constraint violations.
Quantum
40%QUBO validity plus verified circuit evidence, resource efficiency, and route↔solution linkage. Self-reported-only quantum usage is capped at the classical baseline.
Tie-breakers
After total score: fewer constraint violations, shorter total distance, faster execution, earlier submission.
Scoring parameters›
- Baseline route score
- 60
- Constraint violation penalty
- 15
- Average speed (km/h)
- 25
- Self-reported quantum cap
- 40
- QUBO validity points
- 30
- Circuit evidence points
- 40
- Resource efficiency points
- 20
- Route linkage points
- 10
- Declared/verified mismatch penalty
- 15
- Reference qubits
- 16
- Reference circuit depth
- 91
- Reference two-qubit gates
- 168
